[rabbitmq-discuss] Erlang has closed

Fri Sep 18 15:55:47 BST 2009

> Well spotted. It should be 'infinity' instead of 0.0 though, and we
> shouldn't ignore the error if it is something other than timeout.

For my system it was a timeout; as I guess you probably saw :)

>
> I have committed a fix which, subject to qa, will make it into the next
> release.

Is that 1.7, or a 1.6.1 release?

> (Btw, the reason for the timeout is that even though the cpu_sup
> gen_server:call has a timeout of infinity, internally there is a
> measurement_server_call with a timeout of 5s, which presumably gets
> triggered under heavy load).

Ok, that makes sense.  This is probably the biggest wart I've found
with erlang; timeouts are a great way to avoid deadlock when I get
sloppy, but they also break down pretty badly when a machine is under
absurd load and swapping.  I can't think of a better way to handle
things, but in my programs I've gotten pretty tired of things timing
out in weird edge cases where they should just be really slow.